Most Replayed Moment: How AI Companies Are Gaslighting You - Karen Hao

You know I care about data: it lets you act deliberately and turn the lights on in business. But why did Ilya Sutskever leave OpenAI after helping push out Sam Altman?

Ilya felt Altman was steering him toward work he didn’t believe in. He cared about reaching AGI safely and thought Altman’s chaotic, divide-and-conquer management was damaging both.

In 2019, Ilya imagined powerful autonomous AI treating us as we treat animals: not from hatred, but without permission once its priorities take over.

That rests on an unproven view: Ilya and Geoffrey Hinton hypothesize brains are statistical models. Scale a digital model beyond a human brain, and you may get greater intelligence—but many researchers reject that reduction.

But does the mechanism matter if the result is agents doing my work or making my video?

It matters because companies are making world-shaping choices from that hypothesis: consuming data, building data centers, and chasing AGI. Why duplicate and automate people instead of building tools for drug discovery and healthcare?

You interviewed roughly 300 people, including dozens from OpenAI. Why are they pursuing this?

I think it’s an imperial project—hence Empires of AI. These firms claim data and creators’ work, grab land for computing, use contracted labor, and cast themselves as the only ones who can define AI knowledge. They sell concentration as protection from China and call it progress.

When OpenAI shut me out, I worried I’d made a career mistake. But it forced me to report without the front door, and I still conducted more than 300 interviews. What you heard was replayed from an earlier episode; the full conversation is linked below.